Lee Douglas S, Donovan Linda, Austin Peter C, Gong Yanyan, Liu Peter P, Rouleau Jean L, Tu Jack V
Institute for Clinical Evaluative Sciences, Toronto, Canada.
Med Care. 2005 Feb;43(2):182-8. doi: 10.1097/00005650-200502000-00012.
Despite the potential usefulness of administrative databases for evaluating outcomes, coding of heart failure and associated comorbidities have not been definitively compared with clinical data.
To compare the predictive value of heart failure diagnoses and secondary conditions identified in a large administrative database with chart-based records.
The authors studied 1808 patient records sampled from 14 acute care hospitals and compared clinically recorded data with administrative records from the Canadian Institute for Health Information. The impact of comorbidity coding in the administrative data set according to the Charlson classification was examined in models of 30-day mortality.
The positive predictive value (PPV) of a primary diagnosis ICD-9 428 was 94.3% using the Framingham criteria and 88.6% using criteria previously validated with pulmonary capillary wedge pressure. There was reduced prevalence of secondary comorbid conditions in administrative data in comparison with clinical chart data. The specificities and PPV/negative predictive values of administratively identified index comorbidities were high. The sensitivities of index comorbidities were low, but were enhanced by examination of hospitalizations within 1 year prior to the index heart failure admission. Using information from prior hospitalizations modestly enhanced 30-day mortality model performance; however, the odds ratio point estimates of the index and enhanced administrative data sets were consistent with the clinical model.
The ICD-9 428 primary diagnosis is highly predictive of heart failure using clinical criteria. Examination of hospitalization data up to 1 year prior to the index admission improves comorbidity detection and may provide enhancements to future studies of heart failure mortality.
尽管管理数据库在评估预后方面具有潜在用途，但心力衰竭及相关合并症的编码与临床数据尚未得到明确比较。
比较大型管理数据库中确定的心力衰竭诊断和继发性疾病与基于图表记录的预测价值。
作者研究了从14家急症医院抽取的1808例患者记录，并将临床记录数据与加拿大卫生信息研究所的管理记录进行比较。在30天死亡率模型中检查了根据Charlson分类法对管理数据集中合并症编码的影响。
使用弗雷明汉标准，原发性诊断ICD-9 428的阳性预测值（PPV）为94.3%，使用先前经肺毛细血管楔压验证的标准时为88.6%。与临床图表数据相比，管理数据中继发性合并症的患病率较低。管理确定的索引合并症的特异性和PPV/阴性预测值较高。索引合并症的敏感性较低，但通过检查索引心力衰竭入院前1年内的住院情况可提高敏感性。使用先前住院信息适度提高了30天死亡率模型的性能；然而，索引数据集和增强后的管理数据集的优势比点估计与临床模型一致。
使用临床标准，ICD-9 428原发性诊断对心力衰竭具有高度预测性。检查索引入院前1年的住院数据可改善合并症检测，并可能为未来心力衰竭死亡率研究提供改进。
